Before graylog version 6 we had to configure the number of inputbuffer_processors, processbuffer_processors and outputbuffer_processors manually.
Now with the recent version these settings can be removed and graylog could set the settings by itself. However, with the automatic settings I see a drop in message throughput. Out of curiosity I would like to see, how many buffer-processors graylog decided to use… Is there a way to see that?
Side quest:
Can I see the actual CPU Usage per *buffer_processor thread?
I tried to remove these settings to check performances, but it decreases from 11 000 EPS to 5 000 EPS.
So I think auto-detection of CPUs is nice to begin with Graylog but you would have better performances by setting these values.
Regarding the actual number of threads I’ve tried the following commands: